Combinatorics Sum and Product Rules
Problem: How to count without counting.
• How do you figure out how many things there are with
a certain property without actually enumerating all of
them.
Sometimes this requires a lot of cleverness and deep math-
ematical insights.
But there are some standard techniques.
• That’s what we’ll be studying.
1 2
The Sum Rule: If there are n(A) ways to do A and,
distinct from them, n(B) ways to do B, then the number
of ways to do A or B is n(A) + n(B).
• This rule generalizes: there are n(A) + n(B) + n(C)
ways to do A or B or C
• In Section 4.8, we’ll see what happens if the ways of
doing A and B aren’t distinct.
The Product Rule: If there are n(A) ways to do A
and n(B) ways to do B, then the number of ways to do
A and B is n(A) × n(B). This is true if the number of
ways of doing A and B are independent; the number of
choices for doing B is the same regardless of which choice
you made for A.
• Again, this generalizes. There are n(A)×n(B)×n(C)
ways to do A and B and C
3 4
Example 1: In New Hampshire, license plates consisted
of two letters followed by 3 digits. How many possible
license plates are there?
Answer: 26 choices for the first letter, 26 for the second,
10 choices for the first number, the second number, and
the third number:
262 × 103 = 676, 000
Example 2: A traveling salesman wants to do a tour of
all 50 state capitals. How many ways can he do this?
Answer: 50 choices for the first place to visit, 49 for the
second, . . . : 50! altogether.
Chapter 4 gives general techniques for solving counting
problems like this. Two of the most important are:
Some Subtler Examples
Example 3: If there are n Senators on a committee, in
how many ways can a subcommittee be formed?
Two approaches:
1. Let N1 be the number of subcommittees with 1 sen-
ator (n), N2 the number of subcommittees with 2
senator (n(n − 1)/2), . . .
According to the sum rule:
N = N1 + N2 + · · · + N n
n!
• It turns out that Nk = k!(n−k)! (n choose k); this
is discussed in Section 4.4
• A subtlety: What about N0? Do we allow sub-
committees of size 0? How about size n?
◦ The problem is somewhat ambiguous.
If we allow subcommittees of size 0 and n, then
there are 2n subcommittees altogether.
◦ This is the same as the number of subsets of the
set of n Senators: there is a 1-1 correspondence
between subsets and subcommittees.
2. Simpler method: Use the product rule!
 • Each senator is either in the subcommittee or out
of it: 2 possibilities for each senator:
◦ 2 × 2 × · · · × 2 = 2n choices altogether
General moral: In many combinatorial problems, there’s
more than one way to analyze the problem.
5 6
Coping with Ambiguity
If you think a problem is ambiguous:
1. Explain why
2. Choose one way of resolving the ambiguity
3. Solve the problem according to your interpretation
• Make sure that your interpretation doesn’t render
the problem totally trivial
7 8
How many ways can the full committee be split into two
sides on an issue?
This question is also ambiguous.
• If we care about which way each Senator voted, then
the answer is again 2n: Each subcommittee defines
a split + vote (those in the subcommittee vote Yes,
those out vote No); and each split + vote defines de-
fines a subcommittee.
• If we don’t care about which way each Senator voted,
the answer is 2n/2 = 2n−1.
◦ This is an instance of the Division Rule.
More Examples
Example 4: How many legal configurations are there
in Towers of Hanoi with n rings?
Answer: The product rule again: Each ring gets to
“vote” for which pole it’s on.
• Once you’ve decided which rings are on each pole,
their order is determined.
• The total number of configurations is 3n
Example 5: How many distinguishable ways can the
letters of “computer” be arranged? How about “dis-
crete”?
For computer, it’s 8!:
• 8 choices for the first letter, for the second, . . .
Is it 8! for discrete? Not quite.
• There are two e’s
Suppose we called them e1, e2:
• There are two “versions” of each arrangement, de-
pending on which e comes first: discre1te2 is the same
as discre2te1.
• Thus, the right answer is 8!/2!
Division Rule: If there is a k-to-1 correspondence be-
tween of objects of type A with objects of type B, and
there are n(A) objects of type A, then there are n(A)/k
objects of type B.
A k-to-1 correspondence is an onto mapping in which
every B object is the image of exactly k A objects.
Permutations
A permutation of n things taken r at a time, written
P (n, r), is an arrangement in a row of r things, taken
from a set of n distinct things. Order matters.
Example 6: How many permutations are there of 5
things taken 3 at a time?
Answer: 5 choices for the first thing, 4 for the second,
3 for the third: 5 × 4 × 3 = 60.
• If the 5 things are a, b, c, d, e, some possible permuta-
tions are:
abc abd abe acb acd ace
adb adc ade aeb aec aed
...

In general
n!
P (n, r) = = n(n − 1) · · · (n − r + 1)
(n − r)!

9 10

Combinations More Examples

A combination of n things taken r at a time, written Example 8: How many full houses are there in poker?
C(n, r) or nr (“n choose r”) is any subset of r things
 

• A full house has 5 cards, 3 of one kind and 2 of an-

from n things. Order makes no difference. other.
Example 7: How many ways are there of choosing 3 • E.g.: 3 5’s and 2 K’s.
things from 5?
Answer: If order mattered, then it would be 5 × 4 × 3. Answer: You need to find a systematic way of counting:
Since order doesn’t matter, • Choose the denomination for which you have three of
abc, acb, bac, bca, cab, cba a kind: 13 choices.

are all the same. • Choose the three: C(4, 3) = 4 choices

• For way of choosing three elements, there are 3! = 6 • Choose the denomination for which you have two of
ways of ordering them. a kind: 12 choices

Therefore, the right answer is (5 × 4 × 3)/3! = 10: • Choose the two: C(4, 2) = 6 choices.

abc abd abe acd ace Altogether, there are:

ade bcd bce bde cde 13 × 4 × 12 × 6 = 3744 choices
In general
n!
C(n, r) = = n(n − 1) · · · (n − r + 1)/r!
(n − r)!r!

11 12
 0! More Questions

It’s useful to define 0! = 1. Q: How many ways are there of choosing k things from
{1, . . . , n} if 1 and 2 can’t both be chosen? (Suppose
Why?
n, k ≥ 2.)
1. Then we can inductively define
A: First find all the ways of choosing k things from n—
(n + 1)! = (n + 1)n!, C(n, k). Then subtract the number of those ways in
and this definition works even taking 0 as the base which both 1 and 2 are chosen:
case instead of 1. • This amounts to choosing k−2 things from {3, . . . , n}:
2. A better reason: Things work out right for P (n, 0) C(n − 2, k − 2).
and C(n, 0)! Thus, the answer is
How many permutations of n things from n are there? C(n, k) − C(n − 2, k − 2)
n! n!
P (n, n) = = = n! Q: What if order matters?
(n − n)! 0!
A: Have to compute how many ways there are of picking
How many ways are there of choosing n out of n? k things, two of which are 1 and 2.
0 out of n?
  P (n, k) − k(k − 1)P (n − 2, k − 2)
n n!
 = =1




n 
n!0!
 
n n!
 = =1




0  0!n!

13 14

Q: How many ways are there to distribute four distinct Combinatorial Identities
balls evenly between two distinct boxes (two balls go in
each box)?
A: All you need to decide is which balls go in the first There all lots of identities that you can form using C(n, k).
box. They seem mysterious at first, but there’s usually a good
C(4, 2) = 6 reason for them.
Theorem 1: If 0 ≤ k ≤ n, then
Q: What if the boxes are indistinguishable?
C(n, k) = C(n, n − k).
A: C(4, 2)/2 = 3.
Proof:
n! n!
C(n, k) = = = C(n, n−k)
k!(n − k)! (n − k)!(n − (n − k))!

Q: Why should choosing k things out of n be the same

as choosing n − k things out of n?
A: There’s a 1-1 correspondence. For every way of choos-
ing k things out of n, look at the things not chosen: that’s
a way of choosing n − k things out of n.
This is a better way of thinking about Theorem 1 than
the combinatorial proof.

15 16
Theorem 2: If 0 < k < n then
     
n n − 1 n − 1
 =   + 


 

k   k  k − 1

Proof 1: (Combinatorial) Suppose we want to choose k

objects out of {1, . . . , n}. Either we choose the last one
(n) or we don’t.
1. How many ways are there of choosing k without choos-
ing the last one? C(n − 1, k).
2. How many ways are there of choosing k including n?
This means choosing k − 1 out of {1, . . . , n − 1}:
C(n − 1, k − 1).

Proof 2: Algebraic . . .

Note: If we define C(n, k) = 0 for k > n and k < 0,

Theorems 1 and 2 still hold.

17